A Bayesian pressure‑inversion–driven method for establishing mechanically grounded digital twins of in‑service tunnel linings

Accurately sensing the loading status is crucial for the structural health monitoring of in-service tunnel linings. To address this need, this paper proposes a method for constructing a mechanically grounded Digital Twin (DT) of tunnel linings, comprising two components: (i) Bayesian learning of the loading conditions on the linings from easily observable data, and (ii) Driving a mechanical model to replicate the real-world behavior based on these learned conditions. Notably, this method, developed within a statistical framework, systematically quantifies associated uncertainties, thereby providing insights for constructing an informative DT model. A numerical case demonstrates that the developed DT model can accurately reproduce comprehensive structural responses throughout the in-service linings. Furthermore, the model enables virtual trial-and-error simulations for predicting future performance. Additional analyses provide guidance for refining DT models by quantifying uncertainties and identifying effective strategies for their reduction. The proposed method is also validated using two experimentally reported cases from the literature, confirming its effectiveness while also revealing limitations that inform directions for future research.